From: sashiko-bot@kernel.org
To: "Yixun Lan" <dlan@kernel.org>
Cc: linux-scsi@vger.kernel.org, devicetree@vger.kernel.org,
robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH v8 2/2] scsi: ufs: spacemit: k3: Add UFS Host Controller driver
Date: Mon, 16 Nov 2026 04:16:38 +0000 [thread overview]
Message-ID: <sashiko-outbox-216168@kernel.org> (raw)
In-Reply-To: <20261116-08-k3-ufs-support-v8-2-e3b256972dd4@kernel.org>
Thank you for your contribution! Sashiko AI review found 5 potential
issues to consider:
- [High] scsi: ufs: spacemit: clocks not enabled and sys1clk rate always
0 without OPP/freq-table
- [Medium] scsi: ufs: spacemit: zero ref_clk gating delay in
ufs_spacemit_post_hibern8() on UFS 2.x
- [Low] scsi: ufs: spacemit: ufs_spacemit_event_notify() logs cleared
hba->errors on UFS_EVT_ABORT
- [Low] scsi: ufs: spacemit: duplicated ufshcd_init_host_params() and
unused definitions
- [Low] scsi: ufs: spacemit: positive UIC error codes ignored in DME
error checks
--
commit 893661ab0d45a5f00a091213a0890aea0386b457
Author: Yixun Lan <dlan@kernel.org>
scsi: ufs: spacemit: k3: Add UFS Host Controller driver
This commit adds a UFS host controller platform driver for the SpacemiT K3
SoC, supporting JEDEC UFS 2.2, MIPI UniPro v1.61, and M-PHY v3.0.
> diff --git a/drivers/ufs/host/ufs-spacemit.c b/drivers/ufs/host/ufs-spacemit.c
> new file mode 100644
> index 0000000000000..eb2ad50826997
> --- /dev/null
> +++ b/drivers/ufs/host/ufs-spacemit.c
[ ... ]
> +static u32 ufs_spacemit_get_sys1clk_1us(struct ufs_hba *hba)
> +{
> + struct ufs_clk_info *clki, *clka = NULL;
> + struct list_head *head = &hba->clk_list_head;
> + unsigned long rate_hz = 0;
> +
> + if (!list_empty(head)) {
> + list_for_each_entry(clki, head, list) {
> + if (clki->name && !strcmp(clki->name, "axi") && clki->clk) {
> + clka = clki;
> + break;
> + }
> + }
> + }
> +
> + if (clka && clka->clk)
> + rate_hz = clk_get_rate(clka->clk);
> +
> + if (!rate_hz)
> + return 0;
> +
> + return DIV_ROUND_CLOSEST(rate_hz, 1000000);
> +}
[Severity: High]
Does hba->clk_list_head remain empty when the Device Tree node has neither
freq-table-hz nor operating-points-v2?
ufshcd_parse_clock_info() returns early without parsing clock-names when
freq-table-hz is absent, and ufshcd_parse_operating_points() does the same
when operating-points-v2 is absent:
drivers/ufs/host/ufshcd-pltfrm.c:ufshcd_parse_clock_info() {
...
sz = of_property_count_u32_elems(np, "freq-table-hz");
if (sz <= 0) {
dev_info(dev, "freq-table-hz property not specified\n");
goto out;
}
...
}
Since the spacemit,k3-ufshc binding omits both properties and
ufs_spacemit_init() does not populate hba->clk_list_head or acquire the
clocks itself, ufshcd_init_clocks() and ufshcd_setup_clocks() never enable
the axi or ref_clk clocks (leaving them subject to clk_disable_unused or
unclocked MMIO access if not enabled by firmware), hba->dev_ref_clk_freq is
never parsed from ref_clk, and ufs_spacemit_get_sys1clk_1us() always returns
0 so ufs_spacemit_link_startup_pre_change() always falls back to
UFS_AXI_CLK_LOW_FREQ_HZ (409.6 MHz).
[ ... ]
> +static void ufs_spacemit_set_dev_cap(struct ufs_host_params *host_params, u32 pwr_hs)
> +{
> + if (!host_params)
> + return;
> +
> + memset(host_params, 0, sizeof(struct ufs_host_params));
> + host_params->tx_lanes = UFS_SPACEMIT_K3_LIMIT_NUM_LANES_TX;
> + host_params->rx_lanes = UFS_SPACEMIT_K3_LIMIT_NUM_LANES_RX;
> + host_params->hs_rx_gear = UFS_SPACEMIT_K3_LIMIT_HSGEAR_RX;
> + host_params->hs_tx_gear = UFS_SPACEMIT_K3_LIMIT_HSGEAR_TX;
> + host_params->pwm_rx_gear = UFS_SPACEMIT_K3_LIMIT_PWMGEAR_RX;
> + host_params->pwm_tx_gear = UFS_SPACEMIT_K3_LIMIT_PWMGEAR_TX;
> + host_params->rx_pwr_pwm = UFS_SPACEMIT_K3_LIMIT_RX_PWR_PWM;
> + host_params->tx_pwr_pwm = UFS_SPACEMIT_K3_LIMIT_TX_PWR_PWM;
> + host_params->rx_pwr_hs = pwr_hs;
> + host_params->tx_pwr_hs = pwr_hs;
> + host_params->hs_rate = UFS_SPACEMIT_K3_LIMIT_HS_RATE;
> + host_params->desired_working_mode = UFS_HS_MODE;
> +}
[Severity: Low]
This isn't a bug, but ufs_spacemit_set_dev_cap() and the
UFS_SPACEMIT_K3_LIMIT_* macros set the exact same defaults as the exported
ufshcd_init_host_params() helper in drivers/ufs/host/ufshcd-pltfrm.c.
In addition, ufs_spacemit_init() allocates struct ufs_spacemit_host and
stores it with ufshcd_set_variant(), but the variant pointer is never
retrieved anywhere in the driver. MPHY_TX_FSM_STATE is also defined in both
ufs-spacemit.h and ufs-spacemit.c, ANA_EQ_CTRL_REG_ATTR and
MPHY_PU_WITH_HB8_RESET are defined in ufs-spacemit.c while raw literals
0x00CD and 0xb7f are still used in ufs_spacemit_uniprov1p6_init() and
ufs_spacemit_mphy_init(), and MPHY_RX_FSM_STATE along with several headers
and macros in ufs-spacemit.h are unused.
> +
> +static int ufs_spacemit_link_startup_pre_change(struct ufs_hba *hba)
> +{
> + u32 value, sys1clk_1us;
> + int err;
> +
> + err = ufs_spacemit_mphy_init(hba);
> + if (err < 0)
> + return err;
> +
> + err = ufs_spacemit_uniprov1p6_init(hba);
> + if (err < 0)
> + return err;
[ ... ]
> +static int ufs_spacemit_apply_dev_quirks(struct ufs_hba *hba)
> +{
> + static const struct ufshcd_dme_attr_val setup_attrs[] = {
> + /* LCC_DISABLE */
> + { UIC_ARG_MIB_SEL(TX_LCC_ENABLE, UIC_ARG_MPHY_TX_GEN_SEL_INDEX(0)), 0 },
> + { UIC_ARG_MIB_SEL(TX_LCC_ENABLE, UIC_ARG_MPHY_TX_GEN_SEL_INDEX(1)), 0 },
> + /* TX_Min_ActivateTime */
> + { UIC_ARG_MIB_SEL(TX_MIN_ACTIVATETIME, UIC_ARG_MPHY_TX_GEN_SEL_INDEX(0)), 0x0 },
> + { UIC_ARG_MIB_SEL(TX_MIN_ACTIVATETIME, UIC_ARG_MPHY_TX_GEN_SEL_INDEX(1)), 0x0 },
> + { UIC_ARG_MIB(ANA_HSGEAR_CTRL_ATTR), 0x25 },
> + };
> + int err;
> +
> + if (hba->dev_quirks & UFS_DEVICE_QUIRK_HOST_PA_SAVECONFIGTIME) {
> + err = ufs_spacemit_quirk_host_pa_saveconfigtime(hba);
> + if (err < 0)
> + return err;
> + }
> +
> + if (hba->dev_info.wmanufacturerid == UFS_VENDOR_WDC)
> + hba->dev_quirks |= UFS_DEVICE_QUIRK_HOST_PA_TACTIVATE;
> +
> + err = ufs_spacemit_dme_set(hba, setup_attrs, ARRAY_SIZE(setup_attrs));
> + if (err < 0)
> + return err;
> +
> + return ufs_spacemit_wait_mphy_pll_lock(hba);
> +}
[Severity: Low]
As reported by Troy Mitchell on v6 [1], can positive UIC error codes bypass
the "if (err < 0)" checks in ufs_spacemit_link_startup_pre_change() and
ufs_spacemit_apply_dev_quirks()?
ufshcd_dme_get() and ufshcd_dme_set() return positive 8-bit UIC command
result codes (0x01..0xFF) when a DME command fails at the UniPro layer.
Checking "if (err < 0)" misses those positive error codes and allows both
functions to continue past failed DME operations instead of aborting early.
[ ... ]
> +static void ufs_spacemit_event_notify(struct ufs_hba *hba, enum ufs_event_type evt, void *data)
> +{
> + switch (evt) {
> + case UFS_EVT_PA_ERR:
> + if (data)
> + dev_warn(hba->dev, "PA error event, INT errors:0x%x, PA_ERR_CODE:0x%x\n",
> + hba->errors, *(u32 *)data);
> +
> + break;
> + case UFS_EVT_DL_ERR:
> + if (data)
> + dev_warn(hba->dev, "DL error event, INT errors:0x%x, DL_ERR:0x%x\n",
> + hba->errors, *(u32 *)data);
> +
> + break;
> + case UFS_EVT_ABORT:
> + dev_warn(hba->dev, "Abort event, INT errors:0x%x\n", hba->errors);
> + break;
> + default:
> + break;
> + }
> +}
[Severity: Low]
Does the UFS_EVT_ABORT case in ufs_spacemit_event_notify() always log
hba->errors as 0x0?
ufshcd_check_errors() clears hba->errors back to 0 before the interrupt
handler returns (saving sticky bits in hba->saved_err), while UFS_EVT_ABORT
is only emitted later from ufshcd_abort() in SCSI error-handler context with
the aborted tag or LUN passed via data.
[ ... ]
> +static void ufs_spacemit_post_hibern8(struct ufs_hba *hba, enum uic_cmd_dme cmd)
> +{
> + static const struct ufshcd_dme_attr_val setup_attrs[] = {
> + { UIC_ARG_MIB_SEL(0xf1, UIC_ARG_MPHY_RX_GEN_SEL_INDEX(0)), 0x84 },
> + { UIC_ARG_MIB_SEL(0xf1, UIC_ARG_MPHY_RX_GEN_SEL_INDEX(1)), 0x84 },
> + { UIC_ARG_MIB_SEL(0xf1, UIC_ARG_MPHY_RX_GEN_SEL_INDEX(0)), 0x80 },
> + { UIC_ARG_MIB_SEL(0xf1, UIC_ARG_MPHY_RX_GEN_SEL_INDEX(1)), 0x80 },
> + { UIC_ARG_MIB(0xdd), 0x57 },
> + { UIC_ARG_MIB(0xdd), 0xd7 },
> + { UIC_ARG_MIB(0xe8), 0x57 },
> + { UIC_ARG_MIB(0xe8), 0xd7 },
> + };
> + int err;
> +
> + if (cmd == UIC_CMD_DME_HIBER_ENTER) {
> + err = ufs_spacemit_check_hibern8(hba);
> + if (err)
> + dev_err(hba->dev, "ufs_spacemit_check_hibern8 error\n");
> +
> + ufs_spacemit_dme_set(hba, setup_attrs, ARRAY_SIZE(setup_attrs));
> +
> + /* Power down M-PHY */
> + ufshcd_writel(hba, 0x0, UFS_PHY_MNG_BASE + UFS_MPHY_PU_CTRL);
> +
> + fsleep(hba->dev_info.clk_gating_wait_us);
> +
> + /* Disable reference clock, keep device reset deasserted */
> + ufshcd_writel(hba, MPHY_DEVICE_HW_RESET,
> + UFS_PHY_MNG_BASE + UFS_DEVICE_IO_CTRL);
> + }
> +}
[Severity: Medium]
Does fsleep(hba->dev_info.clk_gating_wait_us) gate the device reference
clock with zero delay on UFS 2.x devices?
ufshcd_get_ref_clk_gating_wait() only reads bRefClkGatingWaitTime and sets
hba->dev_info.clk_gating_wait_us when hba->dev_info.wspecversion >= 0x300.
On UFS 2.x devices (or if Hibern8 is entered before
ufshcd_device_params_init() runs), clk_gating_wait_us remains 0, so
ufs_spacemit_post_hibern8() calls fsleep(0) and disables
MPHY_DEVICE_REFCLK_OE immediately after powering down the M-PHY without
satisfying the reference clock hold time.
[1] https://lore.kernel.org/r/DLLQJDEA91FU.18Q1TSX09AUR3@linux.spacemit.com
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261116-08-k3-ufs-support-v8-0-e3b256972dd4@kernel.org?part=2
prev parent reply other threads:[~2026-11-16 4:16 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-16 3:46 [PATCH v8 0/2] Add UFS Host driver support for SpacemiT K3 SoC Yixun Lan
2026-11-16 3:46 ` [PATCH v8 1/2] scsi: ufs: spacemit: dt-bindings: Add UFS controller for " Yixun Lan
2026-11-16 3:46 ` [PATCH v8 2/2] scsi: ufs: spacemit: k3: Add UFS Host Controller driver Yixun Lan
2026-11-16 4:16 ` sashiko-bot [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=sashiko-outbox-216168@kernel.org \
--to=sashiko-bot@kernel.org \
--cc=conor+dt@kernel.org \
--cc=devicetree@vger.kernel.org \
--cc=dlan@kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=robh@kernel.org \
--cc=sashiko-reviews@lists.linux.dev \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox